Speech AI for All: Challenging Deficit-based Approaches to Speech Science and Technology

In recent years, InterSpeech has promoted research agendas around fair and inclusive speech science and technology, seeking to promote research on and by historically marginalized and/or pathologized groups. Despite this focus, current research directions, namely those rooted in diagnostic assessment of “pathological” speech, run the risk of branding affected communities as deficient or abnormal, necessitating both technological surveillance and “correction” of their speech.

Work at InterSpeech on “fair and inclusive” speech science and technology is primarily quantitative in nature, but we believe quantitative approaches alone are insufficient to truly achieve fairness and inclusivity. As other conferences that are concerned with these topics such as FAccT and CHI encourage both quantitative and qualitative work, we propose this special session in the hope of providing a home for work grounded in qualitative and HCI-based methods for system design, analysis, and critique.

This special session is a continuation of many conversations the organizers have had throughout the years both inside and outside of InterSpeech. Most recently at CHI 2025, we hosted the SpeechAI4All workshop, which brought together speech researchers, HCI researchers, and impacted communities to discuss the current and future state of speech technology. These conversations deserve a home within InterSpeech to promote truly fair, accessible, inclusive, and equitable speech science and technology. We cannot solely focus on the technological contributions of our work, but also their downstream, human impacts.